Epilepsy affects from 0.5-1% of the human population, with an estimated 1/3 to nearly 1/2 of patients unable to attain satisfactory seizure control. A majority of patients have seizures of focal onset, possibly with secondary generalization. Within a seizure focus, the behavior of neurons evolves in time and space: this is reflected in the occurrence of low-amplitude, high-frequency (gamma [30-70 Hz], or higher [>70 Hz]) oscillations in a subpopulation of neurons, just prior to initiation of more typical epileptiform waves that then spread. Spatiotemporal evolution has been reproduced in in vitro and in vivo seizure models, as has another nonuniformity: different neuronal firing patterns across cortical layers. An in vitro model of >100 Hz oscillations leading into an electrographic seizure also exists. A better understanding of the detailed pathophysiology of within-focus seizure initiation, followed by spread, is important: a) it might allow improved early warning of an impending seizure, and suggest alternative measures to abort the seizure; b) it could suggest drug treatments, for example binders to proteins constituting putative axonal gap junctions, or modulators of said junctions; c) it could possibly refine the preoperative evaluation of surgical candidates, by indicating the most physiologically based types of preoperative EEG monitoring, for example, concentrating on frequency bands not typically examined. This proposal describes how we can first adapt an existing single-column network model for the study of very fast oscillations and seizure initiation, and then build a multi-column cortical network model. Temporal evolution of a seizure can be studied in one column, and spatial evolution in multiple columns. We use detailed model neuronal elements, along with synaptic and gap junctional connectivity. Such models are necessary as tools for integrating diverse types of experimental measurements (field potentials, single-cell electrical and optical recordings), and for accounting for the contributions to population activity made by a) intrinsic properties of multiple cell types, b) within-column synaptic and electrical connectivity, and c) between-column synaptic connectivity. A successful model will generate specific and testable predictions useful to experimental collaborators, who have worked closely on this project in the past and continue to do so. [unreadable] [unreadable] [unreadable]